PROCESS DEVIATIONS

Control of thermal process operations in food canning factories has consisted of maintaining specified operating conditions that have been predetermined from product and process heat penetration tests, such as the process calculations for the time and temperature of a batch cook. Sometimes unexpected changes can occur during the course of the process operation or at some point upstream in a processing sequence such that the prespecified processing conditions are no longer valid or appropriate, and an off-specification product is produced that must be either reprocessed or destroyed at appreciable economic loss. These types of situations are known as process deviations and can be of critical importance in food processing operations because the physical process variables that can be measured and controlled are often only indicators of complex biochemical reac- tions that take place under the specified process conditions.

Because of the important emphasis placed on the public safety of canned foods, processors operate in strict compliance with the Food and Drug Admin- istration’s low-acid canned food regulations. Among other things, these regula- tions require strict documentation and record keeping of all critical control points in the processing of each retort load or batch of canned product. Particular emphasis is placed on product batches that experience an unscheduled process deviation, such as when a drop in retort temperature occurs during the course of the process, which may result from loss of steam pressure. In such a case, the product will not have received the established scheduled process and must be either destroyed, fully reprocessed, or set aside for evaluation by a competent processing authority. If the product is judged to be safe, then batch records must contain documentation showing how that judgment was reached. If judged unsafe, then the product must be fully reprocessed or destroyed. Such practices are costly. In recent years food engineers knowledgeable in the use of engineering mathematics and scientific principles of heat transfer have developed determin- istic computer models capable of simulating thermal processing of conduction- heated canned foods, such as described in this chapter. These models make use of numerical solutions to mathematical heat transfer equations capable of pre- dicting accurately the internal product cold-spot temperature in response to any dynamic temperature experienced by the retort during the process. As such, they are very useful in the rapid evaluation of deviations that may unexpectedly occur.

Accuracy of such models is of paramount importance, and the models must work equally as well for any mode of heat transfer or size and shape container. Recall that the deterministic model described earlier in this chapter was derived for the case of pure conduction heat transfer in a solid body of finite cylinder shape. It would not be applicable to the many food products that heat by con- vection or to varying degrees of combined convection and conduction, or to different shapes. Recent work reported in the literature has described effective modification and simplification of the model to overcome these limitations.7,13 These reports confirmed that food containers need not be of the same shape as the solid body assumed by the heat transfer model. They could be of any shape so long as temperature predictions were required only at the single cold-spot location within the container from which heat penetration data were determined. The improved model assumed the product was a pure conduction-heating solid in the form of a sphere. An apparent thermal diffusivity was obtained for the solid sphere that would produce the same heating rate as that experienced by the product cold spot. Similarly, the precise radial location where the heating lag

factor (jh) was the same as that at the product cold spot would be used as the location at which temperature would be calculated by the model (Figure 3.15 and Figure 3.16). Thus, for any product with empirical parameters (fh and jh) known from heat penetration tests, it would be possible to simulate the thermal response at the product cold spot to any dynamic boundary condition (time-varying retort temperature) regardless of container size or shape or process conditions (mode of heat transfer).

Recall that heat penetration test data normally produce straight-line semilog heat penetration curves from which the empirical heat penetration parameters (fh and jh) can be determined. Incorporation of the parameters into the heat transfer model is accomplished by the relationship between thermal diffusivity (α) and

heating rate factor (fh) for a sphere (Equation 3.9), and the relationship between

heating lag factor (jh) and radial location (r) within the sphere (Equation 3.10). These and similar relationhips for other regular solid body shapes can be found in the literature.6,13

fh = 0.233 (R2/a) (3.9)

j(r) = 0.637 (R2_{/r) sin(pr/R) (3.10)}

Results from heat penetration tests on five products7 _{are presented in Table 3.5.} All products exhibited straight-line (log-linear) heat penetration curves on semi- logarithmic plots of unaccomplished temperature differences vs. time. Can-to- can variation in the heating rate factor (fh) and lag factors derived from direct analysis of the heat penetration curve ( jh, analyzed) were determined by the maximum and minimum values found over all six cans from two replicate tests. The true heating lag factor found by trial-and-error simulation ( jh, simulated) was also compared. This was the value chosen for use in the heat transfer model along with the maximum fh values (slowest heating) for a conservative routine

simulation of each product. The range of lethality values calculated from the temperatures measured by thermocouples in each can (Fo, actual) were also compared. Lethality was calculated from the simulated temperature profile (Fo, simulated) predicted by the heat transfer model in response to the retort temper- ature data file from each heat penetration test as input. Figure 3.17 compares internal cold-spot temperatures predicted by model simulation with profiles mea- sured by thermocouples in response to multiple retort temperature deviations FIGURE 3.15 Heat penetration curves for five different locations along the radius on the

midplane of a cylindrical container (see Figure 3.16), illustrating relationship between location and heating lag factor (jh).

1 ₂ 3 ₄ 5 20 40 60 80 100 Time (min) 120 119 117 115 113 111 101 81 61 41 21 –79 T emper ature ( °C) ∆t Å Ç É Ñ Ö

during a heat penetration test.7 _{The simulated profiles follow the measured profiles} quite closely in response to relatively severe and twice repeated deviations.

The final test of model performance in the simulation and evaluation of process deviations was a comparison of lethalities accomplished by actual and simulated temperature profiles (Table 3.6). Recall that the accomplished lethality (Fo) for any thermal process is easily calculated by numerical integration of the measured or predicted cold-spot temperature over time, as explained previously. Thus, if the cold-spot temperature can be accurately predicted over time, so can FIGURE 3.16 Replacement of solid body shape from finite cylinder to perfect sphere for

simplification of numerical heat transfer model, with choice of radial location based upon heating lag factor from heat penetration tests.

R r 5 4 3 2 1 h H r R

TABLE 3.5

Heat Penetration Results on Products Using Two Replicated Heat Penetration Tests with Six Instrumented Cans for Each Product

Product and Process fh Range (min) jh Range, Analyzed jh, Simulated Fo Range, Actual Fo, Simulated 5% bentonite, 1-kg cans (98 × 110 mm), static cook 70.4–73.0 1.9–2.0 2.0 6.0–7.0 6.2 5% bentonite, tuna cans (86 × 45 mm), static cook 20.0–22.0 1.4–1.6 1.4 7.5–9.8 7.4 Water, 1-kg cans (98 × 110 mm), static cook 3.0–3.1 1.8–2.3 1.0 9.8–10.8 9.9

Water, tuna cans

(86 × 45 mm), static cook 1.7–1.9 2.5–3.9 1.0 7.9–10.6 7.7 Peas in brine, ∫-kg cans (74 × 88 mm), agitated cook 2.5–3.0 2.6–3.4 1.0 10.8–12.0 11.0

Source: Teixeira, A.A., J. Food Sci., 64, 488–493, 1999. With permission.

FIGURE 3.17 Comparison of internal cold-spot temperatures predicted by model simu-

lation with those measured by thermocouples in response to multiple retort temperature deviations during a heat penetration test with 5% bentonite suspension in a 6-ounce tuna can. (From Teixeira, A.A., J. Food Sci., 64, 488–493, 1999. With permission.)

0 10 20 30 40 50 60 70 80 90 Time (min) TC #7 (Fo min) TC #8 (Fo max) Retort Simulation (fh=22, jh=1.4) T emper ature ( °C) 140 120 100 80 60 40 20 0

accumulated process lethality. In all cases, the simulated lethality predicted agreed most closely with the minimum actual lethality calculated from measured temperature profiles. Model predictions that tend toward the minimum side of the range are always desirable for conservative decision making.